GPU out of memory while training HighResNet but works fine when training Unet.

[Ric1779]

I'm trying to train highresnet but there's an error stating that the gpu is out of memory, but I can train the Unet model. I tried training the highresnet without the residual blocks, i.e. only with three conv layers, but still have the same problem.

[KumoLiu]

Hi @Ric1779, I noticed you are using jupyter, did you try to restart the kernel before your training? Sometimes jupyter can't automatically releasing the cache. Thanks!

[Ric1779]

Yes. I've tried it. But it does not work.

[KumoLiu]

Hi @Ric1779, could you please share a piece of code which can help reproduce the issue, then I could take a deep look at it, thanks!